基于地震属性的支持向量机河道砂体厚度预测

Application of Support Vector Machine to the Prediction of the Thickness of Channel Sand Based on Seismic Attributes

松辽盆地北部中浅层扶余油层的主力储层为河道砂体,以厚度薄、沉积规模小和横向变化快为特点。针对河道砂体的准确识别、砂体厚度的精确预测难题,提出在优化地震属性组合的基础上确定支持向量机模型,进而用支持向量机方法定量预测河道砂体。对振幅、频率、相位、地震波形分类和相干等多种有效地震属性开展优选组合,将最好的组合作为最终输入数据。然后调试SVM关键参数:损失函数C、不敏感损失函数参数ε及γ系数,最终全部井点数据参与计算,得出最后预测结果。预测结果不但能较好地保持地震属性的横向分辨率,整体变化趋势符合研究区的沉积地质规律,而且在井点处吻合程度高,具有较高的砂体厚度定量预测能力。

Main reservoir of Fuyu layer of middle and shallow part of northern Songliao Basin is channel sandy body,which is characterized by its thin thickness,small sedimentary scale and lateral heterogeneity. With regard to this problem,we present the model of support vector machine as a solution. By the optimization of the combination of effective seismic attributes,including amplitude,frequency,phase,seismic waveform classification and coherent,the best group was for the final input data. Then the key parameters of the SVM model include loss function C,insensitive loss function parametersεandγcoefficient,and all of the well point data was involved in the calculation to indicate the thickness of the channel sand. The result can maintain the lateral resolution of the seismic data and reflect its general sedimentary pattern as well. The result of subsequently drilled wells indicated that this method has good qualitatively predictive ability for the thickness of channel sand.